The Role Of AI In Managing Vehicle Data And Simplifying Analytics
Artificial intelligence has many use cases. For example, fleet managers have been using AI for years to gather data about vehicles and drivers. In this article from Forbes, learn more about the challenges and opportunities for AI in the automotive sector.
When we think of artificial intelligence (AI) in automotive, we often jump straight to self-driving cars, a personalized in-vehicle experience and enhanced safety. However, before we can arrive at an automobile that adapts to our immediate needs or handles our daily commute, we need to look at the fuel that feeds the algorithms – vehicle data.
Understanding Vehicle Data Complexity
Over the last decade, data generation in the automotive industry has gone through the roof. In 2014, McKinsey estimated that some connected cars were generating around 25 gigabytes of data per hour, a number that at the time was considered massive. However, research from Lucid Motors (via Visual Capitalist) suggests that modern software-defined vehicles (SDVs) produce anywhere between 1.4 terabytes and 19 terabytes of data an hour. That’s an increase of over 5,000% in just 10 years.
Factors behind this escalation include an increase in sensors (including cameras, lidar and radar), high-performance computing (HPC) clusters and advanced vehicle architectures – all of which have been integrated to accommodate more advanced vehicle capabilities and consumer needs.
The trouble, however, is not the amount of data these systems produce but, rather, how to access and sort the relevant data from the superfluous – a task that can be both challenging and expensive. By leveraging innovative data logging approaches, automakers can eliminate some of the irrelevant data before relaying it to the cloud. However, the true power behind data management and analytics is AI. While the potential use cases for AI-driven data management and analytics are extensive, we will focus on three in particular.
Case Studies For AI-Driven Data Analytics
Fault Analysis
Vehicle faults occur for any number of reasons: manufacturing defects, incorrect software versions, unexpected system glitches, improper maintenance, and normal wear and tear. Naturally, the cause of a fault is not always clear, nor is the solution.
Depending on the fault’s severity, it might be necessary to issue a recall notice – which can cost the automaker hundreds to thousands of dollars per vehicle. Unfortunately, recall scopes are based on a “best guess” in which all vehicles that may be impacted are recalled. This approach is both expensive and time-consuming.
An alternative is to use AI to isolate patterns among vehicles exhibiting faults and drill down their occurrences to determine if they are occurring:
• On a particular software version.
• In a particular geographical region.
• On a specific make or model.
• On vehicles that share a specific attribute.
This not only narrows the scope but expedites fault identification and analysis, saving the automaker both time and money. A 2022 SAE International article discussed this type of use case, in which a method that leveraged AI and machine learning was used to predict a vehicle fault before it happened.
Driver Training
Today, fleet managers and insurance companies base driver scores on how good or bad driving behavior is. However, these scores are often based on a limited number of data points and include no insights for improvement.
AI-driven data analytics can assess driving behaviors based on a wider range of data points and offer vehicle-level personalized recommendations for improvement. Researchers in Beijing experimented with using AI in driver training systems in 2019.
Fostering better driving via real-world insights can not only increase road safety, but it can also lead to optimized vehicle utilization and prolonged operation. In other words, using AI for vehicle data analytics can reduce the total cost of ownership for the fleet manager and insurance premiums for the independent driver. When looking at electric vehicles (EVs) specifically, better driving can also result in a lower battery degradation rate, which in turn increases the vehicle’s ability to maintain its range over time.
Reductive Design
An issue that plagues many EV original equipment manufacturers (OEMs) is battery size. Similar to recalls, EV battery size has traditionally been based on estimations. Automakers make assumptions about the needs of the user and select the battery capacity accordingly. However, by leveraging AI to analyze real-world data such as driving and charging patterns, OEMs can determine whether the battery is oversized, undersized or just right for their target market.
They can also determine how certain owner behaviors impact the life of the battery, such as the distance a vehicle covers between charging cycles, the charge frequency and regularity, and the extent of discharge before charging. OEMs can use this information to make product adjustments and provide guidance to users to help increase battery and vehicle longevity.
Outside of battery sizing, automakers can leverage data to remove underused features. Ford announced in February 2024 that it was removing its driver-assist parallel parking feature based on a data analysis revealing that most customers weren’t using it. By adjusting design based on actual usage information, OEMs can reduce production and vehicle costs, ultimately leading to higher sales.
Potential Challenges
While there is huge potential for AI-driven analytics and vehicle data management, there are still several challenges and limitations that must be addressed.
AI requires massive computational resources to function, including storage, memory and power. It is also susceptible to data privacy concerns such as how user data is protected or how systems are secured against biased algorithms and data. As AI in automotive continues to evolve, we will likely see increased legislation that helps address these issues and ensures protection for drivers and their data.
AI And The Automotive Future
There’s no telling how much data SDVs will be producing 10 years from now, but one thing we do know is that AI will prove instrumental in simplifying vehicle data management and analytics. It has the potential to cut through irrelevant data and provide insightful and actionable insights to improve vehicle safety, efficiency and functionality. One day, we will likely see generative AI creating a more holistic and intuitive driving experience where both driver and vehicle can learn from each other.
This article was written by Mayank Sikaria from Forbes and was legally licensed through the DiveMarketplace by Industry Dive. Please direct all licensing questions to legal@industrydive.com.
